In acute ischemic stroke, prediction of the tissue outcome after reperfusion can be used to identify patients that might benefit from mechanical thrombectomy (MT). The aim of this work was to develop ...a deep learning model that can predict the follow-up infarct location and extent exclusively based on acute single-phase computed tomography angiography (CTA) datasets. In comparison to CT perfusion (CTP), CTA imaging is more widely available, less prone to artifacts, and the established standard of care in acute stroke imaging protocols. Furthermore, recent RCTs have shown that also patients with large established infarctions benefit from MT, which might not have been selected for MT based on CTP core/penumbra mismatch analysis.
All patients with acute large vessel occlusion of the anterior circulation treated at our institution between 12/2015 and 12/2020 were screened (
= 404) and 238 patients undergoing MT with successful reperfusion were included for final analysis. Ground truth infarct lesions were segmented on 24 h follow-up CT scans. Pre-processed CTA images were used as input for a U-Net-based convolutional neural network trained for lesion prediction, enhanced with a spatial and channel-wise squeeze-and-excitation block. Post-processing was applied to remove small predicted lesion components. The model was evaluated using a 5-fold cross-validation and a separate test set with Dice similarity coefficient (DSC) as the primary metric and average volume error as the secondary metric.
The mean ± standard deviation test set DSC over all folds after post-processing was 0.35 ± 0.2 and the mean test set average volume error was 11.5 mL. The performance was relatively uniform across models with the best model according to the DSC achieved a score of 0.37 ± 0.2 after post-processing and the best model in terms of average volume error yielded 3.9 mL.
24 h follow-up infarct prediction using acute CTA imaging exclusively is feasible with DSC measures comparable to results of CTP-based algorithms reported in other studies. The proposed method might pave the way to a wider acceptance, feasibility, and applicability of follow-up infarct prediction based on artificial intelligence.
Background: Non-contrast T1 hypointense infarct cores (ICs) within infarcted myocardium detected using cardiac magnetic resonance imaging (CMR) T1 mapping may help assess the severity of left ...ventricular (LV) injury. However, because the relationship of ICs with chronic LV reverse remodeling (LVRR) is unknown, this study aimed to clarify it.Methods and Results: We enrolled patients with reperfused AMI who underwent baseline CMR on day-7 post-primary percutaneous coronary intervention (n=109) and 12-month follow-up CMR (n=94). Correlations between ICs and chronic LVRR (end-systolic volume decrease ≥15% at 12-month follow-up from baseline CMR) were investigated. We detected 52 (47.7%) ICs on baseline CMR by non-contrast-T1 mapping. LVRR was found in 52.1% of patients with reperfused AMI at 12-month follow-up. Patients with ICs demonstrated higher peak creatine kinase levels, higher B-type natriuretic peptide levels at discharge, lower LV ejection fraction at discharge, and lower incidence of LVRR than those without ICs (26.5% vs. 73.3%, P<0.001) at follow-up. Multivariate logistic regression analysis showed that the presence of ICs was an independent and the strongest negative predictor for LVRR at 12-month follow-up (hazard ratio: 0.087, 95% confidence interval: 0.017–0.459, P=0.004). Peak creatine kinase levels, native T1 values at myocardial edema, and myocardial salvaged indices also correlated with ICs.Conclusions: ICs detected by non-contrast-T1 mapping with 3.0-T CMR were an independent negative predictor of LVRR in patients with reperfused AMI.
Mechanical thrombectomy (MT) has been proved to be a highly effective therapy to treat acute ischemic stroke due to large vessel occlusion. Often, the ischemic core extent on baseline imaging is an ...important determinant for endovascular treatment eligibility. However, computed tomography (CT) perfusion (CTP) or diffusion-weighted imaging may overestimate the infarct core on admission and, consequently, smaller infarct lesions called "ghost infarct cores."
A 4-year-old, previously healthy boy presented with acute-onset, right-sided weakness and aphasia. Fourteen hours after the onset of symptoms, the patient presented with a National Institutes of Health Stroke Scale (NIHSS) score of 22, and magnetic resonance angiography demonstrated a left middle cerebral artery occlusion. MT was not considered because of a large infarct core (infarct core volume: 52 mL; mismatch ratio 1.6 on CTP). However, multiphase CT angiography indicated good collateral circulation, which encouraged MT. Complete recanalization was achieved via MT at 16 hours after the onset of symptoms. The child's hemiparesis improved. Follow-up magnetic resonance imaging was nearly normal and showed that the baseline infarct lesion was reversible, in agreement with neurological improvement (NIHSS score 1).
The selection of pediatric stroke with a delayed time window guided by good collateral circulation at baseline seems safe and efficacious, which suggests a promising value of vascular window.
•Dose-dependent effect of delayed physiotherapy on post-stroke recovery was studied.•Overnight but not 5 h voluntary wheel running promoted post-stroke motor recovery.•Recovery of motor function was ...accompanied by induction of angiogenic proteins.•Induction of angiogenic proteins was observed in peri-infract region and infarct core.•Infarct core maybe involved in mechanisms governing post-stroke functional recovery.
Treatments promoting post-stroke functional recovery continue to be an unmet therapeutic problem with physical rehabilitation being the most reproduced intervention in preclinical and clinical studies. Unfortunately, physiotherapy is typically effective at high intensity and early after stroke – requirements that are hardly attainable by stroke survivors. The aim of this study was to directly evaluate and compare the dose-dependent effect of delayed physical rehabilitation (daily 5 h or overnight voluntary wheel running; initiated on post-stroke day 7 and continuing through day 21) on recovery of motor function in the mouse photothrombotic model of ischemic stroke and correlate it with angiogenic potential of the brain. Our observations indicate that overnight but not 5 h access to running wheels facilitates recovery of motor function in mice in grid-walking test. Western blotting and immunofluorescence microscopy experiments evaluating the expression of angiogenesis-associated proteins VEGFR2, doppel and PDGFRβ in the peri-infarct and corresponding contralateral motor cortices indicate substantial upregulation of these proteins (≥2-fold) in the infarct core and surrounding cerebral cortex in the overnight running mice on post-stroke day 21. These findings indicate that there is a dose-dependent relationship between the extent of voluntary exercise, motor recovery and expression of angiogenesis-associated proteins in this expert-recommended mouse ischemic stroke model. Notably, our observations also point out to enhanced angiogenesis and presence of pericytes within the infarct core region during the chronic phase of stroke, suggesting a potential contribution of this tissue area in the mechanisms governing post-stroke functional recovery.
Stroke is characterised as a cerebrovascular disease, which acts as the key factor of mortality and permanent disabilities. Ischaemic stroke is observed as a widespread stroke that results in tissue ...hypoxia. Hence, it is crucial to analyse the ischaemic stroke lesions for the accurate identification of the stroke in intensive care units. However, the recognition of ischaemic lesions is a difficult process due to their small resolution and poor image resolution. Various schemes have been suggested for stroke lesion identification, localisation, and detection, finding infarct cores and penumbras is of high interest. Artificial intelligence could be one of the promising technologies in all methods, which may accelerate stroke analysis and lead to improved patient recovery. This article provides a review of different articles related to ischaemic lesion deduction using Computed Tomography (CT) and CT perfusion imaging. This review article thus provides insight into methods, performance measures, and the key highlight of these models. Further, this review article provides the challenges encountered in the existing techniques.
Computed tomography perfusion imaging in acute stroke requires further validation. We aimed to establish the optimal computed tomography perfusion parameters defining the infarct core and critically ...hypoperfused tissue. Sub-6-h computed tomography perfusion and 24-h magnetic resonance imaging were analysed from 314 consecutive patients with ischaemic stroke. Diffusion-weighted imaging lesion volume at 24 h was used to define the extent of critically hypoperfused tissue (in patients without reperfusion between acute and 24-h time points), and infarct core (in patients with major reperfusion at 24 h). Pixel-based analysis of co-registered computed tomography perfusion and diffusion-weighted imaging was then used to define the optimum computed tomography perfusion thresholds for critically hypoperfused at-risk tissue and infarct core. These optimized acute computed tomography perfusion threshold-based lesion volumes were then compared with 24-h diffusion-weighted imaging infarct volume, as well as 24-h and 90-day clinical outcomes for validation. Relative delay time >2 s was the most accurate computed tomography perfusion threshold in predicting the extent of critically hypoperfused tissue with both receiver operating curve analysis (area under curve 0.86), and the volumetric validation (mean difference between computed tomography perfusion and 24-h diffusion-weighted imaging lesions = 2 cm2, 95% confidence interval 0.5-3.2 cm2). Cerebral blood flow <40% (of contralateral) within the relative delay time >2 s perfusion lesion was the most accurate computed tomography perfusion threshold at defining infarct core with both receiver operating characteristic analysis (area under curve = 0.85) and the volumetric validation. Using these thresholds, the extent of computed tomography perfusion mismatch tissue (the volume of 'at-risk' tissue between the critically hypoperfused and core thresholds) salvaged from infarction correlated with clinical improvement at 24 h (R
2 = 0.59, P = 0.04) and 90 days (R
2 = 0.42, P = 0.02). Patients with larger baseline computed tomography perfusion infarct core volume (>25 ml) also had poorer recovery at Day 90 (P = 0.039). Computed tomography perfusion can accurately identify critically hypoperfused tissue that progresses to infarction without early reperfusion, and the computed tomography perfusion cerebral blood flow infarct core closely predicts the final volume of infarcted tissue in patients who do reperfuse. The computed tomography perfusion infarct core and at-risk measures identified are also strong predictors of clinical outcome.
Flat detector computed tomography (FD-CT) technology is becoming more widely available in the angiography suites of comprehensive stroke centers. In patients with acute ischemic stroke (AIS), who are ...referred for endovascular therapy (EVT), FD-CT generates cerebral pooled blood volume (PBV) maps, which might help in predicting the final infarct area. We retrospectively analyzed pre- and post-recanalization therapy quantitative PBV measurements in both the infarcted and hypoperfused brain areas of AIS patients referred for EVT.
We included AIS patients with large vessel occlusion in the anterior circulation referred for EVT from primary stroke centers to our comprehensive stroke center. The pre- and post-recanalization FD-CT regional relative PBV (rPBV) values were measured between ipsilateral lesional and contralateral non-lesional areas based on final infarct area on post EVT follow-up cross-sectional imaging. Statistical analysis was performed to identify differences in PBV values between infarcted and non-infarcted, recanalized brain areas.
We included 20 AIS patients. Mean age was 63 years (ranging from 36 to 86 years). The mean pre- EVT rPBV value was 0.57 (±0.40) for infarcted areas and 0.75 (±0.43) for hypoperfusion areas. The mean differences (Δ) between pre- and post-EVT rPBV values for infarcted and hypoperfused areas were respectively 0.69 (±0.59) and 0.69 (±0.90). We found no significant differences (p > 0.05) between pre-EVT rPBV and ΔrPBV values of infarct areas and hypoperfusion areas.
Angiographic PBV mapping is useful for the detection of cerebral perfusion deficits, especially in combination with the fill run images. However, we were not able to distinguish irreversibly infarcted tissue from potentially salvageable, hypoperfused brain tissue based on quantitative PBV measurement in AIS patients.
To explore a new approach mainly based on deep learning residual network (ResNet) to detect infarct cores on non-contrast CT images and improve the accuracy of acute ischemic stroke diagnosis.
We ...continuously enrolled magnetic resonance diffusion weighted image (MR-DWI) confirmed first-episode ischemic stroke patients (onset time: less than 9 h) as well as some normal individuals in this study. They all underwent CT plain scan and MR-DWI scan with same scanning range, layer thickness (4 mm) and interlayer spacing (4 mm) (The time interval between two examinations: less than 4 h). Setting MR-DWI as gold standard of infarct core and using deep learning ResNet combined with a maximum a posteriori probability (MAP) model and a post-processing method to detect the infarct core on non-contrast CT images. After that, we use decision curve analysis (DCA) establishing models to analyze the value of this new method in clinical practice.
116 ischemic stroke patients and 26 normal people were enrolled. 58 patients were allocated into training dataset and 58 were divided into testing dataset along with 26 normal samples. The identification accuracy of our ResNet based approach in detecting the infarct core on non-contrast CT is 75.9%. The DCA shows that this deep learning method is capable of improving the net benefit of ischemic stroke patients.
Our deep learning residual network assisted with optimization methods is able to detect early infarct core on non-contrast CT images and has the potential to help physicians improve diagnostic accuracy in acute ischemic stroke patients.
Introduction: In patients with acute ischemic stroke, the location and volume of an irreversible infarct core determine prognosis and treatment. We aimed to determine if automated CT perfusion (CTP) ...is non-inferior to diffusion-weighted imaging (DWI) or fluid-attenuated inversion recovery (FLAIR) in predicting the acute infarct core. Methods: In this systematic review and meta-analysis, we searched MEDLINE and EMBASE from 1960 to December 2020. Five outcome measures were examined: volumetric difference, volumetric correlation, sensitivity and specificity at the patient level, Dice coefficient, and sensitivity and specificity at the voxel level. A random-effects meta-analysis was performed for volumetric difference and correlation. Results: From 3,986 studies retrieved, 48 studies met our inclusion criteria with 46 studies on anterior circulation, one study on posterior circulation, and one study on lacunar infarct strokes. In anterior circulation stroke, there were no significant mean volumetric differences between CTP and acute DWI (cerebral blood flow CBF 0.52 mL, 95% CI −0.07, 1.11, I 2 0.0%; relative CBF rCBF 3.01 mL, 95% CI −0.46, 6.48, I 2 82.6%; relative cerebral blood volume rCBV −12.84 mL, 95% CI −38.56, 12.88, I 2 96.2%) and between CTP and delayed DWI or FLAIR (rCBF −1.29 mL, 95% CI −6.49, 3.92, I 2 91.8%; rCBV −5.80 mL, 95% CI −16.20, 4.60, I 2 84.2%). Mean correlation between CTP and acute DWI was 0.90 (95% CI 0.80, 0.95, I 2 60.0%) for rCBF and 0.84 (95% CI 0.58, 0.94, I 2 93.5%) for rCBV. Mean correlation between CTP and delayed DWI or FLAIR was 0.74 (95% CI 0.57, 0.85, I 2 94.6%) for rCBF and 0.90 (95% CI 0.69, 0.97, I 2 93.1%) for rCBV. Sensitivity and specificity at the patient level were reported by three studies and Dice coefficient by four studies. Statistical analysis could not be performed for sensitivity and specificity at the voxel level. Limited evidence was available for posterior circulation or lacunar infarct strokes. Conclusion: Due to significant heterogeneity and insufficient high-quality studies reporting each outcome, there is insufficient evidence to reliably determine the accuracy of CTP prediction of the infarct core compared to DWI or FLAIR.
The benefits of endovascular treatment (EVT) on large ischemic infarct core mainly focus on a core size of 70–150 ml. The relationship between EVT and very large ischemic infarct core (>150 ml) is ...unclear. We herein present an acute stroke patient who achieved functional independence after EVT without postoperative decompressive craniectomy despite very large ischemic infarct core.
A 50-year-old Asian male was admitted to our hospital with “sudden disturbance of consciousness with left limb weakness for 11 hours”. The patient had a history of clipping treatment for ruptured aneurysms. After an emergency CTA and CTP, very large ischemic core of 190 ml and a mismatch ratio (Tmax > 6s volume/core volume) of 1.9 were shown in preoperative imaging. EVT was performed, and postoperative strict monitoring was conducted without decompressive craniectomy. The patient was discharged from the hospital on the 16th day, scoring 2 on the modified Rankin scale at a 2-year follow-up.
Imaging suggests very large ischemic infarct core; if there is a substantial mismatch between major functional areas (large ischemic penumbra) and the patient is relatively young, aggressive EVT may be beneficial.